Konstantin Georgiev (MSc, BEng (Hons))

Thesis title: Predicting rehabilitation needs and trajectories of older patients

Precision Medicine DTP

Year of study: 3

  • Precision Medicine DTP
  • BHF/University Centre for Cardiovascular Science
  • College of Medicine & Veterinary Medicine

Contact details

Address

Street

Chancellor's Building
49 Little France Crescent
Little France Campus
The Royal Infirmary of Edinburgh

City
Edinburgh
Post code
EH164SB

Availability

  • I'm generally flexible and open to opportunities at the moment, feel free to get in touch by email.

Background

A third-year PhD student and Health Data Scientist, specialising in developing clinical decision-support tools. I use predictive analytics to provide health experts with actionable insights for care pathways in primary and secondary care services. Possessing a fair amount of experience in collaboration with health and care providers across NHS Scotland, I aim to provide solutions that can contribute to better quality of life for burdened patients and better resource management in care settings.

Qualifications

MSc graduate in Artificial Intelligence (University of Aberdeen), MSc DataLab alumni

BEng graduate in Software Engineering (Technical University of Varna)

Responsibilities & affiliations

Precision Medicine DTP student

Health Data Scientist at RedStar AI (https://redstar.ai/)

HighSTEACS, Ageing and Health, AIAI (School of Informatics) research groups

Research summary

My current interests lie in the field of Population Health Data Science, more specifically, analysis of care pathways, risk prediction and resource management in complex patients. This includes, but is not limited to older, multimorbid, frail groups and COVID-19 related conditions. Throughout the years, I’ve been primarily involved with incorporating Electronic Health Record (EHR) data collected from hospitals and emergency care services for determining risk trajectories and unmet clinical needs. The current work of my PhD is focused on developing an EHR-based risk score for rehabilitation complexity in older patients, estimated through novel NHS Lothian data describing received multidisciplinary and out-of-hours specialist care. As a Health Data Scientist, I use Machine Learning techniques (e.g. Ensemble Learning) to predict hospital mortality, readmission and adverse events in time-sensitive cohorts (e.g. diabetes and hip fracture patients). Additionally, I have some experience in the field of Explainable AI (XAI), specifically for adapting Process Mining techniques to explore care patterns and abnormal behaviour in sub-populations, as well as identifying important model predictors.

Project activity

Thesis title: ”Predicting rehabilitation needs and trajectories of older patients

Abstract:

Frailty and acute disability are growing concerns within healthcare services, many outcomes of which are unpredictable. Healthcare resources that can be allocated to support the rehabilitation of individuals with complex symptoms are finite. Additionally, inactivity during hospitalisation often results in increased risk of decline in independence, quality-of-life and can even lead to the acquisition of new disabilities. Untimely intervention and unclear decision boundaries that define the patient condition are typically the cause of this. Current approaches in geriatric medicine do not support personalised care for multimorbid patients. With the recent introduction of Electronic Health Records, there is a major unmet potential, in terms of data-driven approaches for rehabilitation assessment. These data can be utilised to build structured care pathways by identifying key factors that drive improvement or decline in patient health throughout their recovery process and ensure the approach that minimises the patient risk is undertaken. These analyses can then be communicated to professionals using explainable Machine Learning techniques to allow individualised patient profiling and treatment management.

Previous and Current work:

The initial research was focused on a statistical analysis of the relationships between COVID-19 status and rehabilitation provision (out-of-hours, specialised nursing and therapy minutes per week) in unscheduled admissions. I used Process Mining techniques (Inductive Miner with Petri Nets) on timestamped activities with care providers (e.g. nurses, dieticians, physiotherapists) to evaluate the change in patterns between Wave 1 and 2 of the UK pandemic. I used Conformance Checking measures (token-based replay) to quantitatively evaluate the similarity between the mined Petri Nets representing the Wave 1 and 2 cohorts. This allowed me to establish some of the confounding factors during the outbreak period. The current work utilises NHS Lothian hospital data from a range of sources (demographics, lab tests, prescribing, comorbidity history, ward movements and adverse events) to develop and later predict a cumulative score describing rehabilitation complexity at point of admission. Further work will aim to visualise mined patient pathways through a decision-support dashboard with flexible prediction timepoints up to hospital discharge. These visualisations can then be communicated to health and care professionals to aid in patient-level profiling and resource management.

Current project grants

Sir Jules Thorn PhD Scholarship programme
Health Improvement Scotland research grant for the project "Developing an artificial intelligence tool for dementia risk from routine healthcare data" in collaboration with Red Star AI, The Usher Institute

Past project grants

The MSc DataLab Scholarship award

Conference details

"COMPARING CARE PATHWAYS BETWEEN COVID-19 PANDEMIC WAVES USING ELECTRONIC HEALTH RECORDS: A PROCESS MINING CASE STUDY" (2023)

This subgroup analysis utilises Process Discovery (Inductive Mining) techniques to mine Petri Nets that describe hospital care patterns during Wave 1 and 2 of the UK pandemic, within NHS Lothian. Methods for Conformance Checking (precision and fitness) and graph similarity (Graph Edit Distance) are adapted to measure differences in care provision between Wave 1 and 2 for complex cases (e.g. multimorbid, out-of-hours care groups). Preliminary findings linked to rehabilitation care were presented as a poster at the Process-oriented Data Science for Healthcare (PODS4H) workshop at the International Conference for Process Mining (ICPM 2023).

URL: https://pods4h.com/https://drive.google.com/file/d/1fXD_fMI2z7Q3MBV9BU6IrTXx5dBrn5dl/view

"WHAT DEFINES REHABILITATION FOR PATIENTS WITH COVID-19? AN OVERVIEW OF STUDIES IN THE COCHRANE REH-COVER SERIES" (2023)

Description: Brief report and statistical analysis summarising definitions of delivered rehabilitation care on a global scale during the COVID-19 pandemic. This summary is based on the COCHRANE Rapid-living Systematic Review Series (REH-COVER). Related poster was presented at the British Geriatric Society (BGS) Spring Meeting 2023.

URL: To be published in Age and Ageing

"UNDERSTANDING QUANTITY AND INTENSITY OF HOSPITAL REHABILITATION USING ELECTRONIC HEALTH RECORD DATA" (2022)

Description: Statistical analysis describing the impact of COVID-19 status on rehabilitation complexity, describing novel routinely-collected measures for intensity of hospital care. Preliminary findings were presented as an oral communication at the European Geriatric Medicine Society conference (EUGMS 2022, Abstract: O-028).

URL: https://link.springer.com/article/10.1007/s41999-022-00711-8#change-history